second; since the default hz has changed to 1000 times a second,
this resulted in unecessary work being performed.
MFC after: 2 weeks
Discussed with: phk, cperciva
General head nod: silby
tcpip_fillheaders()
tcp_discardcb()
tcp_close()
tcp_notify()
tcp_new_isn()
tcp_xmit_bandwidth_limit()
Fix a locking comment in tcp_twstart(): the pcbinfo will be locked (and
is asserted).
MFC after: 2 weeks
structure, so assert the inpcb lock associated with the tcptw.
Also assert the tcbinfo lock, as tcp_timewait() may call
tcp_twclose() or tcp_2msl_rest(), which require it. Since
tcp_timewait() is already called with that lock from tcp_input(),
this doesn't change current locking, merely documents reasons for
it.
In tcp_twstart(), assert the tcbinfo lock, as tcp_timer_2msl_rest()
is called, which requires that lock.
In tcp_twclose(), assert the tcbinfo lock, as tcp_timer_2msl_stop()
is called, which requires that lock.
Document the locking strategy for the time wait queues in tcp_timer.c,
which consists of protecting the time wait queues in the same manner
as the tcbinfo structure (using the tcbinfo lock).
In tcp_timer_2msl_reset(), assert the tcbinfo lock, as the time wait
queues are modified.
In tcp_timer_2msl_stop(), assert the tcbinfo lock, as the time wait
queues may be modified.
In tcp_timer_2msl_tw(), assert the tcbinfo lock, as the time wait
queues may be modified.
MFC after: 2 weeks
on the tcpcb, but also calls into tcp_close() and tcp_twrespond().
Annotate that tcp_twrecycleable() requires the inpcb lock because it does
a series of non-atomic reads of the tcpcb, but is currently called
without the inpcb lock by the caller. This is a bug.
Assert the inpcb lock in tcp_twclose() as it performs a read-modify-write
of the timewait structure/inpcb, and calls in_pcbdetach() which requires
the lock.
Assert the inpcb lock in tcp_twrespond(), as it performs multiple
non-atomic reads of the tcptw and inpcb structures, as well as calling
mac_create_mbuf_from_inpcb(), tcpip_fillheaders(), which require the
inpcb lock.
MFC after: 2 weeks
protects access to the ISN state variables.
Acquire the tcbinfo write lock in tcp_isn_tick() to synchronize
timer-driven isn bumping.
Staticize internal ISN variables since they're not used outside of
tcp_subr.c.
MFC after: 2 weeks
A complete rationale and discussion is given in this message
and the resulting discussion:
http://docs.freebsd.org/cgi/mid.cgi?4177C8AD.6060706
Note that this commit removes only the functional part of T/TCP
from the tcp_* related functions in the kernel. Other features
introduced with RFC1644 are left intact (socket layer changes,
sendmsg(2) on connection oriented protocols) and are meant to
be reused by a simpler and less intrusive reimplemention of the
previous T/TCP functionality.
Discussed on: -arch
(sorele()/sotryfree()):
- This permits the caller to acquire the accept mutex before the socket
mutex, avoiding sofree() having to drop the socket mutex and re-order,
which could lead to races permitting more than one thread to enter
sofree() after a socket is ready to be free'd.
- This also covers clearing of the so_pcb weak socket reference from
the protocol to the socket, preventing races in clearing and
evaluation of the reference such that sofree() might be called more
than once on the same socket.
This appears to close a race I was able to easily trigger by repeatedly
opening and resetting TCP connections to a host, in which the
tcp_close() code called as a result of the RST raced with the close()
of the accepted socket in the user process resulting in simultaneous
attempts to de-allocate the same socket. The new locking increases
the overhead for operations that may potentially free the socket, so we
will want to revise the synchronization strategy here as we normalize
the reference counting model for sockets. The use of the accept mutex
in freeing of sockets that are not listen sockets is primarily
motivated by the potential need to remove the socket from the
incomplete connection queue on its parent (listen) socket, so cleaning
up the reference model here may allow us to substantially weaken the
synchronization requirements.
RELENG_5_3 candidate.
MFC after: 3 days
Reviewed by: dwhite
Discussed with: gnn, dwhite, green
Reported by: Marc UBM Bocklet <ubm at u-boot-man dot de>
Reported by: Vlad <marchenko at gmail dot com>
to control the packets injected while in sack recovery (for both
retransmissions and new data).
- Cleanups to the sack codepaths in tcp_output.c and tcp_sack.c.
- Add a new sysctl (net.inet.tcp.sack.initburst) that controls the
number of sack retransmissions done upon initiation of sack recovery.
Submitted by: Mohan Srinivasan <mohans@yahoo-inc.com>
- Trailing tab/space cleanup
- Remove spurious spaces between or before tabs
This change avoids touching files that Andre likely has in his working
set for PFIL hooks changes for IPFW/DUMMYNET.
Approved by: re (scottl)
Submitted by: Xin LI <delphij@frontfree.net>
for structures with timers in them. It might be that a timer might fire
even when the associated structure has already been free'd. Having type-
stable storage in this case is beneficial for graceful failure handling and
debugging.
Discussed with: bosko, tegge, rwatson
structures, allowing in6_pcbnotify() to lock the pcbinfo and each
inpcb that it notifies of ICMPv6 events. This prevents inpcb
assertions from firing when IPv6 generates and delievers event
notifications for inpcbs.
Reported by: kuriyama
Tested by: kuriyama
somewhat clearer, but more importantly allows for a consistent naming
scheme for suser_cred flags.
The old name is still defined, but will be removed in a few days (unless I
hear any complaints...)
Discussed with: rwatson, scottl
Requested by: jhb
originated on RELENG_4 and was ported to -CURRENT.
The scoreboarding code was obtained from OpenBSD, and many
of the remaining changes were inspired by OpenBSD, but not
taken directly from there.
You can enable/disable sack using net.inet.tcp.do_sack. You can
also limit the number of sack holes that all senders can have in
the scoreboard with net.inet.tcp.sackhole_limit.
Reviewed by: gnn
Obtained from: Yahoo! (Mohan Srinivasan, Jayanth Vijayaraghavan)
reference count:
- Assert SOCK_LOCK(so) macros that directly manipulate so_count:
soref(), sorele().
- Assert SOCK_LOCK(so) in macros/functions that rely on the state of
so_count: sofree(), sotryfree().
- Acquire SOCK_LOCK(so) before calling these functions or macros in
various contexts in the stack, both at the socket and protocol
layers.
- In some cases, perform soisdisconnected() before sotryfree(), as
this could result in frobbing of a non-present socket if
sotryfree() actually frees the socket.
- Note that sofree()/sotryfree() will release the socket lock even if
they don't free the socket.
Submitted by: sam
Sponsored by: FreeBSD Foundation
Obtained from: BSD/OS
labeling new mbufs created from sockets/inpcbs in IPv4. This helps avoid
the need for socket layer locking in the lower level network paths
where inpcb locks are already frequently held where needed. In
particular:
- Use the inpcb for label instead of socket in raw_append().
- Use the inpcb for label instead of socket in tcp_output().
- Use the inpcb for label instead of socket in tcp_respond().
- Use the inpcb for label instead of socket in tcp_twrespond().
- Use the inpcb for label instead of socket in syncache_respond().
While here, modify tcp_respond() to avoid assigning NULL to a stack
variable and centralize assertions about the inpcb when inp is
assigned.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, McAfee Research
TIME_WAIT recycling cases I was able to generate with http testing tools.
In short, as the old algorithm relied on ticks to create the time offset
component of an ISN, two connections with the exact same host, port pair
that were generated between timer ticks would have the exact same sequence
number. As a result, the second connection would fail to pass the TIME_WAIT
check on the server side, and the SYN would never be acknowledged.
I've "fixed" this by adding random positive increments to the time component
between clock ticks so that ISNs will *always* be increasing, no matter how
quickly the port is recycled.
Except in such contrived benchmarking situations, this problem should never
come up in normal usage... until networks get faster.
No MFC planned, 4.x is missing other optimizations that are needed to even
create the situation in which such quick port recycling will occur.
the syscall arguments and does the suser() permission check, and
kern_mlock(), which does the resource limit checking and calls
vm_map_wire(). Split munlock() in a similar way.
Enable the RLIMIT_MEMLOCK checking code in kern_mlock().
Replace calls to vslock() and vsunlock() in the sysctl code with
calls to kern_mlock() and kern_munlock() so that the sysctl code
will obey the wired memory limits.
Nuke the vslock() and vsunlock() implementations, which are no
longer used.
Add a member to struct sysctl_req to track the amount of memory
that is wired to handle the request.
Modify sysctl_wire_old_buffer() to return an error if its call to
kern_mlock() fails. Only wire the minimum of the length specified
in the sysctl request and the length specified in its argument list.
It is recommended that sysctl handlers that use sysctl_wire_old_buffer()
should specify reasonable estimates for the amount of data they
want to return so that only the minimum amount of memory is wired
no matter what length has been specified by the request.
Modify the callers of sysctl_wire_old_buffer() to look for the
error return.
Modify sysctl_old_user to obey the wired buffer length and clean up
its implementation.
Reviewed by: bms
amount of segments it will hold.
The following tuneables and sysctls control the behaviour of the tcp
segment reassembly queue:
net.inet.tcp.reass.maxsegments (loader tuneable)
specifies the maximum number of segments all tcp reassemly queues can
hold (defaults to 1/16 of nmbclusters).
net.inet.tcp.reass.maxqlen
specifies the maximum number of segments any individual tcp session queue
can hold (defaults to 48).
net.inet.tcp.reass.cursegments (readonly)
counts the number of segments currently in all reassembly queues.
net.inet.tcp.reass.overflows (readonly)
counts how often either the global or local queue limit has been reached.
Tested by: bms, silby
Reviewed by: bms, silby
This is the first of two commits; bringing in the kernel support first.
This can be enabled by compiling a kernel with options TCP_SIGNATURE
and FAST_IPSEC.
For the uninitiated, this is a TCP option which provides for a means of
authenticating TCP sessions which came into being before IPSEC. It is
still relevant today, however, as it is used by many commercial router
vendors, particularly with BGP, and as such has become a requirement for
interconnect at many major Internet points of presence.
Several parts of the TCP and IP headers, including the segment payload,
are digested with MD5, including a shared secret. The PF_KEY interface
is used to manage the secrets using security associations in the SADB.
There is a limitation here in that as there is no way to map a TCP flow
per-port back to an SPI without polluting tcpcb or using the SPD; the
code to do the latter is unstable at this time. Therefore this code only
supports per-host keying granularity.
Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6),
TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective
users of this feature, this will not pose any problem.
This implementation is output-only; that is, the option is honoured when
responding to a host initiating a TCP session, but no effort is made
[yet] to authenticate inbound traffic. This is, however, sufficient to
interwork with Cisco equipment.
Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with
local patches. Patches for tcpdump to validate TCP-MD5 sessions are also
available from me upon request.
Sponsored by: sentex.net
resource exhaustion attacks.
For network link optimization TCP can adjust its MSS and thus
packet size according to the observed path MTU. This is done
dynamically based on feedback from the remote host and network
components along the packet path. This information can be
abused to pretend an extremely low path MTU.
The resource exhaustion works in two ways:
o during tcp connection setup the advertized local MSS is
exchanged between the endpoints. The remote endpoint can
set this arbitrarily low (except for a minimum MTU of 64
octets enforced in the BSD code). When the local host is
sending data it is forced to send many small IP packets
instead of a large one.
For example instead of the normal TCP payload size of 1448
it forces TCP payload size of 12 (MTU 64) and thus we have
a 120 times increase in workload and packets. On fast links
this quickly saturates the local CPU and may also hit pps
processing limites of network components along the path.
This type of attack is particularly effective for servers
where the attacker can download large files (WWW and FTP).
We mitigate it by enforcing a minimum MTU settable by sysctl
net.inet.tcp.minmss defaulting to 256 octets.
o the local host is reveiving data on a TCP connection from
the remote host. The local host has no control over the
packet size the remote host is sending. The remote host
may chose to do what is described in the first attack and
send the data in packets with an TCP payload of at least
one byte. For each packet the tcp_input() function will
be entered, the packet is processed and a sowakeup() is
signalled to the connected process.
For example an attack with 2 Mbit/s gives 4716 packets per
second and the same amount of sowakeup()s to the process
(and context switches).
This type of attack is particularly effective for servers
where the attacker can upload large amounts of data.
Normally this is the case with WWW server where large POSTs
can be made.
We mitigate this by calculating the average MSS payload per
second. If it goes below 'net.inet.tcp.minmss' and the pps
rate is above 'net.inet.tcp.minmssoverload' defaulting to
1000 this particular TCP connection is resetted and dropped.
MITRE CVE: CAN-2004-0002
Reviewed by: sam (mentor)
MFC after: 1 day
rfc3042 Limited retransmit
rfc3390 Increasing TCP's initial congestion Window
inflight TCP inflight bandwidth limiting
All my production server have it enabled and there have been no
issues. I am confident about having them on by default and it gives
us better overall TCP performance.
Reviewed by: sam (mentor)
wait, rather than the socket label. This avoids reaching up to
the socket layer during connection close, which requires locking
changes. To do this, introduce MAC Framework entry point
mac_create_mbuf_from_inpcb(), which is called from tcp_twrespond()
instead of calling mac_create_mbuf_from_socket() or
mac_create_mbuf_netlayer(). Introduce MAC Policy entry point
mpo_create_mbuf_from_inpcb(), and implementations for various
policies, which generally just copy label data from the inpcb to
the mbuf. Assert the inpcb lock in the entry point since we
require consistency for the inpcb label reference.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
zeroed. Doing a bzero on the entire struct route is not more
expensive than assigning NULL to ro.ro_rt and bzero of ro.ro_dst.
Reviewed by: sam (mentor)
Approved by: re (scottl)
